Logic-Regularized Verifier Elicits Reasoning from LLMs
On the map: LLM reasoning → Reasoning benchmarks and evaluation
Abstract (arXiv)
Verifiers are crucial components for enhancing modern LLMs' reasoning capability. Typicalverifiers require resource-intensive superviseddataset construction, which is costly and faceslimitations in data diversity. In this paper, wepropose LOVER, an unsupervised verifier regularized by logical rules. LOVER treats theverifier as a binary latent variable, utilizinginternal activations and enforcing three logical constraints on multiple reasoning paths:negation consistency, intra-group consistency,and inter-group consistency (grouped by thefinal answer). By incorporating logical rulesas priors, LOVER can leverage unlabeled examples and is directly compatible with any offthe-shelf LLMs. Experiments on 10 datasetsdemonstrate that LOVER significantly outperforms unsupervised baselines, achieving performance comparable to the supervised verifier(reaching its 95% level on average). The sourcecode is publicly available at https://github.com/wangxinyufighting/llm-lover.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 10 of its 41 references that are in the map. First posted on arXiv 7 May 2026.
Builds on
10 of 41 references.
References · 10
- 2024The Llama 3 Herd of Modelscited by 14,715
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,754
- 2023Mistral 7Bcited by 3,164
- 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
- 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,807
- 2016Understanding intermediate layers using linear classifier probescited by 1,353
- 2024Large Language Models: A Surveycited by 525
- 2024A Survey on Knowledge Distillation of Large Language Modelscited by 284
- 2020LogiQA: A Challenge Dataset for Machine Reading Comprehension with Logical Reasoningcited by 69
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Verifying the Verifiers: Unveiling Pitfalls and Potentials in Fact Verifiers
- 2026Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning
- 2025LogicReward: Incentivizing LLM Reasoning via Step-Wise Logical Supervision
- 2024Improving LLM Reasoning through Scaling Inference Computation with Collaborative Verification
- 2023DetermLR: Augmenting LLM-based Logical Reasoning from Indeterminacy to Determinacy
- 2025SATQuest: A Verifier for Logical Reasoning Evaluation and Reinforcement Fine-Tuning of LLMs
- 2025Lightweight Latent Verifiers for Efficient Meta-Generation Strategies
- 2024Leveraging LLMs for Hypothetical Deduction in Logical Inference: A Neuro-Symbolic Approach